Color-coated sheet surface defect detection method and device based on unsupervised learning

The invention is suitable for the technical field of color-coated sheet detection, and provides a color-coated sheet surface defect detection method and device based on unsupervised learning. The method comprises the following steps: acquiring a to-be-recognized image of a color-coated sheet through a linear array camera; an effective area is extracted from the to-be-recognized image, and the color-coated sheet pattern type of the effective area is recognized through a color-coated sheet pattern classification model; according to the pattern type of the color-coated sheet and the effective area, performing color-coated sheet defect detection by using a defect detection model, and outputting a defect area coordinate; and extracting a defect area from the to-be-identified image based on the defect area coordinate, and inputting the defect area into a defect type classification and grading model to obtain the defect type and level of the color-coated sheet. Compared with a traditional method based on image featu